library(DiagrammeR)
library(dplyr)
##
## Attaching package: 'dplyr'
## The following objects are masked from 'package:stats':
##
## filter, lag
## The following objects are masked from 'package:base':
##
## intersect, setdiff, setequal, union
library(tidyr)
library(ggplot2)
require(hhi)
## Loading required package: hhi
require(data.table)
## Loading required package: data.table
##
## Attaching package: 'data.table'
## The following objects are masked from 'package:dplyr':
##
## between, first, last
theme_set(theme_bw())
all <- read.csv("data/csss-all.csv")
dakota <- read.csv("data/csss-dakota.csv")
emily <- read.csv("data/csss-emily.csv")
fabian <- read.csv("data/csss-fabian.csv")
jackie <- read.csv("data/csss-jackie.csv")
kyle <- read.csv("data/csss-kyle.csv")
generate_csss_network_df <- function(df, iter){
partic.df <- df %>%
filter(iteration == as.factor(iter)) %>%
select(Name:Location) %>%
unique() %>%
mutate(
Name = as.factor(Name),
name.id = as.numeric(droplevels(Name)))
edge.df.temp <- df %>%
filter(iteration == iter) %>%
select(Title, name1 = Name, Year) %>%
left_join(df, by = c("Title", "Year")) %>%
filter(name1 != Name) %>%
select(Title, name1, name2 = Name, Topic, Topic_isced, Year) %>%
arrange(name1) %>%
mutate(name1 = as.factor(name1),
name2 = as.factor(name2),
name.id1 = as.numeric(droplevels(name1)),
name.id2 = as.numeric(droplevels(name2)),
name.link = paste(name.id1, name.id2, sep = "_"))
edge.df <- edge.df.temp %>%
select(name.id1, name.id2) %>%
mutate(name.link = ifelse(name.id1<name.id2, paste(name.id1, name.id2, sep = "_"), paste(name.id2, name.id1, sep = "_"))) %>%
select(name.link) %>%
unique() %>%
left_join(edge.df.temp, by = "name.link")
nodes.df <- create_node_df(
n = nrow(partic.df),
nodes = partic.df$name.id,
label = partic.df$Name,
color = "red",
year = iter,
discp = partic.df$Discipline_isced,
nat = partic.df$Nationality,
pos = partic.df$Position,
cntry = partic.df$Country_University
)
edges.df <- create_edge_df(
from = edge.df$name.id1,
to = edge.df$name.id2,
dir = rep("none", length(edge.df$name.id1)))
graph <- create_graph(nodes_df = nodes.df,
edges_df = edges.df)
return(graph)
}
graph_network <- function(graph_code) {
render_graph(graph_code)
}
generate_university_network_df <- function(df, year = NA, colorByContinent = FALSE){
if (!is.na(year)){
df <- filter(df, Year == year)
}
euro.countries <- c("Austria", "Belgium", "Czech Republic", "Denmark", "Estonia", "France", "Germany",
"Hungary", "Iceland", "Ireland", "Italy", "Netherlands", "Norway", "Poland", "Portugal", "Russia", "Serbia",
"Serbia and Montenegro", "Slovenia", "Spain", "Sweden", "Switzerland", "The Netherlands", "Turkey", "UK", "United Kingdom")
afri.countries <- c("Nigeria", "South Africa")
asian.countries <- c("China", "India", "Iran", "Israel", "Japan", "Lebanon", "Hong Kong", "Pakistan", "South Korea", "Singapore", "Signapore")
s.amer.countries <- c("Argentina", "Argentinia", "Brazil", "Colombia", "Columbia", "Chile", "Venezuela")
n.amer.countries <- c("Canada", "Canda", "Costa Rica", "Guatemala", "Mexico", "United States", "US", "USA")
univ.df <- df %>%
select(Name, Affiliation:Location) %>%
filter(!is.na(Affiliation)) %>%
unique() %>%
mutate(
Affiliation = as.factor(Affiliation),
aff.id = as.numeric(droplevels(Affiliation))
) %>%
group_by(Affiliation, aff.id) %>%
summarise(aff.count = n()) %>%
ungroup() %>%
left_join(unique(select(df, Affiliation, Country_University)), by = "Affiliation") %>%
mutate(Country_University = as.character(Country_University),
ContColor = ifelse(Country_University %in% euro.countries, "blue",
ifelse(Country_University %in% afri.countries, "green",
ifelse(Country_University %in% asian.countries, "red",
ifelse(Country_University %in% s.amer.countries, "orange",
ifelse(Country_University %in% n.amer.countries, "black", "purple"))))))
temp.edge.df <- df %>%
select(Title, aff1 = Affiliation, name1 = Name, Year) %>%
left_join(df, by = c("Title", "Year")) %>%
filter(aff1 != Affiliation, !is.na(aff1), !is.na(Affiliation)) %>%
select(Title, name1, name2 = Name, aff1, aff2 = Affiliation, Topic_isced, Country_University) %>%
mutate(aff1 = as.factor(aff1),
aff2 = as.factor(aff2),
aff1.id = as.numeric(droplevels(aff1)),
aff2.id = as.numeric(droplevels(aff2)),
title.id = as.numeric(droplevels(as.factor(Title))),
aff.link = paste(title.id, aff1.id, aff2.id, sep = "_"))
univ.edge.df <- temp.edge.df %>%
select(title.id, aff1.id, aff2.id) %>%
mutate(aff.link = ifelse(aff1.id<aff2.id,
paste(title.id, aff1.id, aff2.id, sep = "_"),
paste(title.id, aff2.id, aff1.id, sep = "_"))) %>%
select(aff.link) %>%
unique() %>%
left_join(temp.edge.df, by = "aff.link") %>%
group_by(aff1.id, aff2.id) %>%
mutate(aff.link.count = n()) %>%
ungroup()
if (colorByContinent){
nodes.df <- create_node_df(
n = nrow(univ.df),
nodes = univ.df$aff.id,
#label = univ.df$Affiliation,
fillcolor = univ.df$ContColor,
#height = univ.df$aff.count
)
}else{
nodes.df <- create_node_df(
n = nrow(univ.df),
nodes = univ.df$aff.id,
#label = univ.df$Affiliation,
#height = univ.df$aff.count
)
}
edges.df <- create_edge_df(
from = univ.edge.df$aff1.id,
to = univ.edge.df$aff2.id,
dir = rep("none", nrow(univ.edge.df)))
graph <- create_graph(nodes_df = nodes.df,
edges_df = edges.df) %>%
add_global_graph_attrs(
attr = "layout",
value = "neato",
attr_type = "graph")
return(graph)
}
render_graph(generate_university_network_df(all, year = 2017, colorByContinent = TRUE))
generate_all_graphs <- function(df) {
data.list <- list()
for(i in 1:length(unique(df$iteration))) {
data.list[[i]] <- generate_csss_network_df(df, iter = unique(df$iteration)[i])
}
return(data.list)
}
all.graphs <- generate_all_graphs(all)
#averages for entire network by year
graph_measures <- function(graph) {
df <- data.frame(graph = get_node_df(graph)$year[1],
avg_degree = NA,
max_ecct = NA,
avg_btwn = NA,
avg_path = NA,
avg_const = NA)
df$avg_degree <- get_agg_degree_total(graph, "mean")
df$max_ecct <- get_max_eccentricity(graph)
df$avg_btwn <- mean(get_betweenness(graph)$betweenness)
df$avg_path <- get_mean_distance(graph)
df$avg_const <- mean(get_constraint(graph)$constraint)
return(df)
}
yearly_compare <- function(graphs) {
compile <- graph_measures(graphs[[1]])
for(i in 2:length(graphs)) {
compile[nrow(compile) + 1,] <- graph_measures(graphs[[i]])
}
return(compile)
}
yearly.compare <- yearly_compare(all.graphs)
avdg <- ggplot(data = yearly.compare) +
geom_col(mapping = aes(x = graph, y = avg_degree)) +
coord_flip() +
labs(y = "Average Degree", x = "Year and Location")
diam <- ggplot(data = yearly.compare) +
geom_col(mapping = aes(x = graph, y = max_ecct)) +
coord_flip() +
labs(y = "Diameter", x = "Year and Location")
path <- ggplot(data = yearly.compare) +
geom_col(mapping = aes(x = graph, y = avg_path)) +
coord_flip() +
labs(y = "Average Path Length", x = "Year and Location")
cons <- ggplot(data = yearly.compare) +
geom_col(mapping = aes(x = graph, y = avg_const)) +
coord_flip() +
labs(y = "Average Constraint", x = "Year and Location")
cowplot::plot_grid(avdg, diam, path, cons)
#degree distribution
palette <- palette(rainbow(17))
dd.plot <- ggplot()
data <- rbindlist(lapply(1:length(all.graphs), function(x) {
df <- get_degree_distribution(all.graphs[[x]])
df$network_id <- as.factor(x)
return(df)
}
))
data %>%
ggplot(aes(x = degree, y = total_degree_dist, color = network_id)) +
geom_line()
#community detection comparison
compare_communities <- function(graph) {
gn.com <- get_cmty_edge_btwns(graph)
wt.com <- get_cmty_walktrap(graph)
return(cor(gn.com$edge_btwns_group, wt.com$walktrap_group))
}
create_group_graph <- function(graph, detection = "gn") {
groups <- get_node_df(graph)
if(detection == "gn") {
groups <- left_join(groups, get_cmty_edge_btwns(graph), by = "id")
} else {
groups <- left_join(groups, get_cmty_walktrap(graph), by = "id")
}
if(!require(RColorBrewer)) {install.packages("RColorBrewer")}
palette <- palette(rainbow(length(unique(groups[,ncol(groups)]))))
groups$color <- palette[groups[,ncol(groups)]]
return(create_graph(groups, get_edge_df(graph)))
}
compare_communities(all.graphs[[2]])
## [1] 0.729994
gn.graph <- create_group_graph(all.graphs[[2]], detection = "gn")
wt.graph <- create_group_graph(all.graphs[[2]], detection = "wt")
#what do we want to see?
#
#node-measures
node_measures <- function(graph) {
nodes <- get_node_df(graph)
results <- nodes %>%
left_join(get_degree_total(graph), by = "id") %>%
left_join(get_betweenness(graph), by = "id") %>%
left_join(get_eigen_centrality(graph), by = "id") %>%
left_join(get_constraint(graph), by = "id")
results$discp_hmph <- NULL
results$pos_hmph <- NULL
results$cntry_hmph <- NULL
#attempting to get homophily measures --need to add gender homophily, and prestige homophily
edges <- get_edge_df(graph)
for(i in unique(nodes$id)) {
cur.edges <- edges[edges$from == i,]
cur.edges <- cur.edges[-1]
colnames(cur.edges) <- c("from", "id", "rel", "dir")
cur.edges <- left_join(cur.edges, nodes, by = "id")
if(!is.null(dim(cur.edges))) {
discp.shares <- as.data.frame(table(cur.edges$discp))[as.data.frame(table(cur.edges$discp))$Freq != 0, ]
discp.shares$prop <- discp.shares$Freq/sum(discp.shares$Freq)
results$discp_hmph[results$id == i] <- hhi(discp.shares, "prop")
pos.shares <- as.data.frame(table(cur.edges$pos))[as.data.frame(table(cur.edges$pos))$Freq != 0, ]
pos.shares$prop <- pos.shares$Freq/sum(pos.shares$Freq)
results$pos_hmph[results$id == i] <- hhi(pos.shares, "prop")
cntry.shares <- as.data.frame(table(cur.edges$cntry))[as.data.frame(table(cur.edges$cntry))$Freq != 0, ]
cntry.shares$prop <- cntry.shares$Freq/sum(cntry.shares$Freq)
results$cntry_hmph[results$id == i] <- hhi(cntry.shares, "prop")
}
}
return(results)
}
##need to get data for all graphs
get_all_node_data <- function(graphs) {
compile <- list()
for(i in 1:length(graphs)) {
compile[[i]] <- node_measures(graphs[[i]])
}
return(compile)
}
node.data <- suppressWarnings(get_all_node_data(all.graphs))
#summing measures by discipline
graph_discp_data <- function(nd) {
discp.groups <- nd %>%
group_by(discp) %>%
summarize(dg = mean(total_degree),
btwn = mean(betweenness),
eigen = mean(eigen_centrality),
cnst = mean(constraint),
discp_hmph = mean(discp_hmph),
pos_hmph = mean(pos_hmph),
cntry_hmph = mean(cntry_hmph))
discp.groups$discp <- stringr::str_replace(discp.groups$discp, "[,]", ";")
split <- discp.groups %>%
separate(discp, c("discp1", "discp2", "discp3"), sep = ";", remove = TRUE, fill = "right") %>%
gather(key, DISCP, discp1, discp2, discp3) %>%
filter(!is.na(DISCP)) %>%
group_by(DISCP) %>%
summarize(dg = mean(dg),
btwn = mean(btwn),
eigen = mean(eigen),
cnst = mean(cnst),
discp_hmph = mean(discp_hmph),
pos_hmph = mean(pos_hmph),
cntry_hmph = mean(cntry_hmph))
split %>%
gather(key, value, dg:eigen, discp_hmph:cntry_hmph) %>%
ggplot(aes(x = DISCP, y = value)) +
geom_col() +
coord_flip() +
facet_wrap(~key, scale = "free_x") +
labs(title = nd$year[1])
}
for(i in 5:length(node.data)) {
print(graph_discp_data(node.data[[i]]))
}
# node.data %>%
# gather(key, value, total_degree, betweenness, eigen_centrality, discp_hmph, pos_hmph, cntry_hmph) %>%
# ggplot(aes(x = value)) +
# geom_histogram() +
# facet_wrap(~key, scale = "free_x")